Covalency in the 4f Shell of tris-Cyclopentadienyl ... - ACS Publications

Nov 4, 2011 - pubs.acs.org/JACS. Covalency in the 4f Shell of tris-Cyclopentadienyl Ytterbium. (YbCp3)—A Spectroscopic Evaluation. Robert G. Denning...
0 downloads 0 Views 4MB Size
ARTICLE pubs.acs.org/JACS

Covalency in the 4f Shell of tris-Cyclopentadienyl Ytterbium (YbCp3)—A Spectroscopic Evaluation Robert G. Denning,*,† Jeffrey Harmer,‡ Jennifer C. Green,† and Mark Irwin† †

Department of Chemistry and ‡Center for Advanced Electron Spin Resonance, Department of Chemistry, University of Oxford, South Parks Road, Oxford OX1 3QR, United Kingdom

bS Supporting Information ABSTRACT: Evidence is presented of significant covalency in the ytterbium 4f shell of tris-cyclopentadienyl ytterbium (YbCp3) in its electronic ground state, that can be represented by the superposition of an ionic configuration Yb(III):4f13(Cp3) and a charge-transfer configuration Yb(II):4f14(Cp3)1. The relative weights of these configurations were determined from (i) the difference in their 4f photoionization cross sections, (ii) the accumulation of spin-density centered on the 13C atoms of the Cp ring, as measured by a pulsed EPR (HYSCORE) experiment, (iii) the reduction in the spin-density in the 4f shell, manifest in the 171Yb hyperfine interaction, and (iv) the principal values of the g-tensor, obtained from the EPR spectrum of a frozen glass solution at 5.4 K. Each of these methods finds that the spin density attributable to the charge transfer configuration is in the range 1217%. The presence of configuration interaction (CI) also accounts for the highly anomalous energies, intensities, and vibronic structure in the “ff” region of the optical spectrum, as well as the strict adherence of the magnetic susceptibility to the Curie law in the range 30300 K.

1. INTRODUCTION The high electropositivity of the lanthanides, together with the contracted and screened character of the 4f valence shell, is often taken to imply that, in the ubiquitous III oxidation state, covalency is insignificant and the bonding is primarily electrostatic. However, electron-rich and strongly basic ligands, such as cyclopentadienyl (Cp) and bis(trimethylsilyl)amido anions, form volatile compounds, soluble in nonpolar noncoordinating solvents, in which ligand-to-metal charge-transfer must be substantial. The photoelectron spectroscopy of CeCp31 and LuCp32 taken together with theoretical work3 indicates that interactions with the lanthanide 5d shell are the primary source of this covalency. In the 4f shell any covalency is more difficult to quantify, but it is often taken to be negligible. However, the energy-selective photoelectron spectrum (PES) of YbCp3 strongly suggests that the ground state can be described as an admixture of the Yb(III):4f13L and Yb(II):4f14L1 configurations, where L refers to the ligand molecular orbitals (MO) collectively.2 With the exception of EuCp3 (for which E1/2 = 1.54 V, relative to the SHE, in THF solution), YbCp3 has a less negative reduction potential (E1/2 = 1.92 V) than other LnCp3,4 so configurations involving charge-transfer from electron-rich ligands to the ytterbium atom should be most apparent in this compound. There is, for example, strong magnetic and spectroscopic evidence in compounds of the type Cp*2Yb(bipy) that both the |f14(bipy)æ and |f13(bipy)æ configurations are represented in the ground state.57 r 2011 American Chemical Society

In this paper we present experimental evidence from (i) our previous PES data,2 (ii) new EPR measurements, and (iii) existing electronic spectroscopic studies8,9 that allows a semiquantitative determination of the extent of this class of configuration interaction (CI) in YbCp3. The strikingly anomalous electronic spectrum of YbCp3 in solution,9 and excellent quality spectra in the solid state at low temperatures,8 are long-standing observations that have, for the most part, remained uninterpreted. In what follows, we demonstrate that the high symmetry of this compound, together with the presence of Kramers degeneracy, makes the description of the main features of its electronic structure, and the interpretation of its physical properties, relatively straightforward. As an aside, we note that routine computational methods have not proved satisfactory in addressing this problem. For example, two different DFT codes give values for the spin density on the Yb atom of 39 and 57%. Apart from their mutual inconsistency, when set against our experimental results, both methods seriously overestimate the extent of spin transfer onto the ligands. Despite this inadequacy, DFT calculations have proved useful in establishing certain characteristics of the electronic structure.

2. METHODS 2.1. Crystal Growth. YbCp3 was prepared as described previously.2 A concentrated toluene solution in a Schlenk tube, under argon, was Received: October 3, 2011 Published: November 04, 2011 20644

dx.doi.org/10.1021/ja209311g | J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

ARTICLE

Figure 1. Structure of YbCp3 at 105 K, (a) emphasizing the nominal C3 axis, and (b) showing nonbonded inter-ring interactions. placed in a 40 C freezer. After 2 days, green blocklike crystals of YbCp3, suitable for single-crystal X-ray diffraction, had formed. 2.2. X-ray Crystallography. Data were collected using an EnrafNonius Kappa-CCD diffractometer and a 95 mm CCD area detector with a graphite-monochromated molybdenum Kα source (λ = 0.71073 Å). Crystals were selected under Paratone-N oil, mounted on MiTeGen loops, and slowly cooled to 105 K, using an open flow N2 cooling device, at a rate of 1 K per minute.10 Data were processed using the DENZO-SMN package, including unit cell parameter refinement and interframe scaling (using SCALEPACK within DENZO-SMN).11 Structures were subsequently solved using direct methods, and refined on F2 (R1 = 0.0194, wR2 = 0.0447) using the SHELXL 972 package.12,13 2.3. Electron Paramagnetic Resonance. X-band CW EPR experiments were performed on a Bruker BioSpin GmbH EMX spectrometer equipped with a high sensitivity Bruker probehead, and pulse EPR experiments were carried out at X-band on a Bruker E680 spectrometer using a 4 mm dielectric ring resonator (EN 4118X-MD4). Both instruments were equipped with Oxford Instruments helium-flow cryostats. All pulse data were recorded at 9.7512 GHz and 3.75 K. The EPR spectrum was recorded by integrating the FID created with an 800 ns microwave (mw) pulse. Hyperfine sublevel correlation (HYSCORE)14,15 experiments employed the pulse sequence π/2τπ/2t1πt2π/ 2τecho using the following parameters: mw pulses of lengths tπ/2 = tπ = 16 ns, starting times of 96 ns for t1 and t2, and time increments Δt = 16 ns (data matrix 400  400) and τ = 114 ns. An eight-step phase cycle was used to remove unwanted echoes. The HYSCORE data were processed with MATLAB 7.0 (The MathWorks, Inc.). The time traces were baseline corrected with an exponential, apodized with a Gaussian window, and zero filled. After a two-dimensional Fourier transformation, absolute-value spectra were calculated. The CW EPR and HYSCORE spectra were simulated with the program EasySpin.16 2.4. Raman Spectroscopy. A solid sample was flame-sealed in a Pyrex capillary under dinitrogen. The spectrum was recorded on a Dilor Labram 300 spectrometer, using 20 mW HeNe laser excitation, at 632.8 nm. To minimize photodecomposition, this laser power was attenuated by a factor of 100. The counting time was 120 s. The spectrometer was calibrated before each measurement by reference to the 520.7 nm line of a Si wafer. 2.5. Computation. Density functional calculations were carried out using the Amsterdam Density Functional program suite ADF 2010.2.1719 Scalar relativistic corrections were included via the ZORA method for all calculations.2023 The generalized gradient

approximation was employed, using the local density approximation of Vosko, Wilk, and Nusair24 together with the nonlocal exchange correction by Becke25,26 and nonlocal correlation corrections by Perdew.27 TZP basis sets were used with triple-ζ accuracy sets of Slater type orbitals and polarization functions added to the main group atoms. The cores of the atoms were frozen up to 1s for C and 4d for Yb. Default conditions were used for the optimizations and frequency calculations. Calculations were also performed using the Gaussian program suite.28 The latter employed the B3LYP functional24,2931 and SDD basis sets with a StuttgartBonn relativistic effective core potential for Yb.

3. RESULTS 3.1. Structure and Geometry of YbCp3. The single crystal X-ray structure of YbCp3 was first reported32 in 1987. One of the Cp rings was found to be disordered, so the structure was refined using D5h symmetry and fixed bond lengths in that ring. The structure has now been redetermined at 105 K; it is effectively unchanged but is more precise, and there is no evidence of disorder. The molecular symmetry is close to C3h (Figure 1a), with the planes of the Cp rings being almost exactly parallel to the pseudo3-fold axis; they are, however, tilted about axes parallel to C3 so that the YbC5 distances are ∼0.1 Å longer than YbC2 and YbC3. A space-filling model implies that this arises from steric interactions between the rings and that the ring orientations may be constrained by the interlocking of the C5/H5 atoms in the notional C3h plane, with C2/H2 and C3/H3 on adjacent rings (Figure 1b). Selected bond lengths and angles are listed in the Supporting Information. The distances from the Yb atom to the centroids of the rings are 2.359(2), 2.343(2), and 2.357(2) Å. The Yb atom is displaced by 0.055 Å along the normal to the plane formed by the ring centroids, which is substantially larger than the offset of 0.010 Å from the centroid of centroids within that plane. The centroidYbcentroid angles are 119.40(7), 120.47(7), and 119.96(7). The shortest intermolecular contact is Yb(1)C4C(2) = 4.042(4) Å (cf. packing diagram in Figure S1 of the Supporting Information). However, there is no significant difference between the Yb(1)C4C(1) and Yb(1)C1C(1) distances (cf. table in the Supporting Information), which are locally related by 20645

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

ARTICLE

gii

^

)

Table 1. Spin-Hamiltonian Parameters for YbCp3 in a Frozen Glass at 5.25 K

1.9245

6.0659

173

354

1322

171

1287

4805

Yb, Aii/MHz Yb, Aii/MHz

Figure 2. X-band CW EPR spectrum of YbCp3 at 5.25 K (red), and simulated (black).

the nominal C3h plane, so intermolecular interactions do not appear to be electronically significant. Equivalent distances in the pyrazine-bridged adduct Cp3Yb(μpyrazine)YbCp3 are longer, at 2.43(1), 2.38(1), and 2.42(1) Å, and the average of the NYbCp(centroid) angles is 98.5.33 A similar flattened tetrahedral geometry is found in the triphenylphosphine oxide adduct, in which the OYbCp(centroid) angles average 100,34 as well as in LnCp3 ammonia35 and THF36 adducts, where these angles are ∼98.5. In these adducts, the Yb offset from the plane of the ring centroids is typically 0.35 Å. 3.2. Electron Paramagnetic Resonance. 3.2.1. CW X-Band EPR in Frozen Glass Solution. The 9.386 GHz spectrum of a ∼1 mM frozen glass solution of YbCp3 in an equimolar mixture of benzene, toluene, and heptane at 5.25 K is shown Figure 2. The features between 110 mT and 120 mT in Figure 2a, marked (i), (ii), and (iii), imply that three species are present. Component (i), whose spectrum can be fitted to an axial spin Hamiltonian, accounts for 77% of the intensity. Components (ii) and (iii) constitute 15.7% and 7.7% of the intensity, respectively, with a g-tensor that appears to be orthorhombic in both cases. The satellites, seen clearly in Figure 2a, are due to the 171Yb (I = 1/2) and 173Yb (I = 5/2) hyperfine interactions. These nuclei have natural abundances of 14.4% and 16.2%, respectively. The differences between the principal values of the g-tensor for all three species are small, implying that variations in the electronic and geometrical structure are also small. The axially symmetric g-tensor of component (i) is consistent with the C3h symmetry expected in the noncoordinating solvent. The lower symmetry of species (ii) and (iii) is, we believe, due to the

Figure 3. (a) FID-detected EPR spectrum of YbCp3 (black) and its numerical first derivative (blue). (b) HYSCORE recorded near g^ (see arrow in part a) showing the 13C region: experiment (top) and simulation (bottom) using As = 0.4 and Ap = 1.6 MHz. 20646

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

)

trapping of different relative ring orientations within the glass. A simulated spectrum using the spin-Hamiltonian parameters in Table S8 of the Supporting Information is shown in black in Figure 2. Agreement with experiment is good in the low-field region and also above 335 mT, but it is less good between 315 and 335 mT. In the latter region, there are too many unconstrained variables associated with components (ii) and (iii) for further refinement to be useful. Only the properties at the C3h site (i), for which the parameters (Table 1) are well-defined, merit further discussion. 3.2.2. Pulsed X-Band EPR. Figure 3a shows the absorption spectrum at 9.75 GHz, detected by integration of the freeinduction decay of the magnetization following a selective 800 ns microwave pulse. The numerical derivative of the absorption spectrum is comparable to that in Figure 2. However, due to the short T2 time of molecules with g aligned parallel to the static magnetic field, the resonances near 110 mT are not detected (the spectrometer dead-time is ∼100 ns). HYSCORE measurements14 were made at 359 mT and at 300 mT, at which fields the 13C nuclear resonance frequencies are 3.844 and 3.212 MHz, respectively. At 359 mT a group of molecules, oriented with their C3 axes within a small solid angle Table 2. Selected Vibrational Frequencies expt (cm1) C3h D5h mode no.37 D5h irrep

description

112

E0

CpYbCp in plane bend

147

A0

sym CYbC asym stretch

196

E0

asym CYbC asym stretch

215 241

A00 A0

sym ring tilt about axis in plane YbCp sym stretch

255

E0

619

E0

14

E200

CCC out of plane deformation

846

E0

12

E20

CCC in plane bend

1065

E0

11

E20

CH in plane bend

1356

E0

10

E20

CC in plane stretch

YbCp asym stretch

ARTICLE

that is close to perpendicular to the static field, are in resonance with the microwave field. HYSCORE spectra correlate a nuclear frequency in the α electron spin manifold with a nuclear frequency in the β electron spin manifold. In the weak coupling case (2|ν13C| < |A|) crosspeaks occur in the (+,+) quadrant, as is the case in Figure 3b. The hyperfine couplings are approximately given by the splitting around |ν13C| as indicated, and anisotropy in the hyperfine matrix is proportional to the maximum shift of the cross-peaks behind the antidiagonal. The simulation shown is the sum of five 13 C hyperfine couplings from a Cp ring and is discussed in section 4.5. 3.3. Vibrational Spectroscopy. The frequencies, in the ground electronic state, of those modes that give rise to vibronic structure in the electronic absorption spectrum are summarized in Table 2. They come from several sources, including the Raman spectrum measured in this work, and can be assigned by using the predictions of a DFT calculation, along with data from a survey of Cp compounds.37 They remain constant to within ∼5 cm1 across a series of LnCp3 (Ln = La, Pr, Nd, Sm, Gd, Dy, Tm).38 The full data set is in Table S2 and Figure S2 of the Supporting Information. 3.4. Electronic Spectroscopy. 3.4.1. Data Sources. The absorption and emission spectra of YbCp3 in benzene solution, together with those of several adducts, were reported in 1983 by Schlesener and Ellis.9 In 1967 the absorption spectrum of solid YbCp3, sublimed onto a quartz window, was measured by Pappalardo and Jørgensen8 at nominal temperatures of 4.2, 77, and 298 K. Neither publication contains any significant analysis of the spectral details. Both groups used the legendary Cary 14 spectrophotometer, and both sets of diagrams appear to have been reproduced photographically from the instrument charts. Where necessary, we have used a graphics editor to separate overlaid traces in the original diagrams and also to remove a weak sharp band in the solution spectra at 985 nm attributed9 to a decomposition product. Edited traces were then digitized automatically (Digitizeit: www.digitizeit.de).

Figure 4. Spectrum of a 5.5 mM solution of YbCp3 in benzenepath-length 1 cm, after Schlesener et al.9 Inset: the spectra of YbCp3 (black), Yb(dpa)3 3 13H2O at 15 K (red), and KY3F10 at 298 K (blue). 20647

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

ARTICLE

Table 3. Integrated Intensities in the 2F7/2 f 2F5/2 Region compd YbCp3 (9,00013,500 cm1) Yb(dpa)3 3 13H2O, at 15 K39

oscillator strength/106 182 5.1

KY3F10 at 298 K40

3.1

Yb3+(aqueous) at 298 K41

3.3

Figure 6. Spectrum at 4.2 K (red) and 298 K (black). Shifts to lower energy with increasing temperature are shown in red, and those to higher energy are in black.

Figure 5. Spectrum of YbCp3 at 298 K in benzene solution, from Schlesener et al.9 (blue), and in the solid state, from Pappalardo and Jørgensen8 (black).

The solid sample was mounted in an evacuated glass cell in an immersion cryostat, but its temperature was not measured.8 The paper states that low noise spectra in the near-infrared were taken in short time intervals, immediately after the liquid He level fell below the light path. The long time constants needed to acquire high resolution low-noise spectra in this region (where the responsivity of the PbS detector is poor) suggest that the data were acquired in short spectral segments, during each of which the vacuum-mounted sample would be susceptible to warming. Anomalies and discontinuities in the spectra appear to indicate that several such segments have been concatenated. A strongly sloping baseline in the low temperature spectra of the solid between 500 and 650 nm is absent at room temperature, implying that scattering increases sharply at shorter wavelengths, possibly due to microscopic fracturing of the “as deposited” glassy film on cooling. A subjective baseline correction was therefore applied in order to establish comparability with the room temperature spectrum.

3.4.2. General Characteristics. The spectrum of YbCp3 in benzene solution is shown in Figure 4 together with the spectra of two conventional Yb3+ compounds with known absorption cross sections, Yb(dpa)3 3 13H2O (dpa = 2,6-pyridinedicarboxylate) at 15 K (shown in red)39 and Yb3+ doped into a crystal of KY3F10 at room temperature (in blue).40 The structure in the range 900013,500 cm1, a region in which the 2F7/2 f 2F5/2 transitions of Yb3+ would normally be expected and which in the reference compounds spans 1000 cm1 and thus accounts for the accurate Curie law dependence of the susceptibility between 20 and 300 K.2 4.4. The 171Yb Hyperfine Interaction. The dipolar part of this interaction is a linear function of the expectation value Æri3æ, where ri is the electronnuclear distance. For an electron localized in the 4f orbitals of a free Yb3+ ion, Æri3æ is calculated to be 12.85 Bohr3.53 If the ground state is derived exclusively from the 2F7/2 free ion state, the ratio of the principal values of the dipolar part of the A tensor to those of the g tensor should be a constant.54 Figure 12 demonstrates this using 46 sets of 171Yb EPR data from 28 different Yb3+compounds (section S6 of the Supporting Information), most of which are oxides or fluorides with cubic or axial site symmetry and are expected to have the properties of the 4f13 ion. The slope of this plot gives Akk/gkk = 796 ( 2.5 MHz. The intercept of 43 ( 3 MHz in Figure 12 is attributable to the isotropic (contact) hyperfine coupling and agrees adequately with the value of 21 ( 17 MHz reported many years ago.54 The scatter in Figure 12 is due to small second-order contributions

20654

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

ARTICLE

C1

C2

C3

C4

C5

total

p-orbitals

0.27

1.26

1.29

0.29

2.26

4.25 ( 0.27

s-orbitals total

0.02 0.29

0.03 1.28

0.03 1.32

0.02 0.31

0.06 2.32

0.07 ( 0.02 4.32 ( 0.30

4.5. Carbon-13 Hyperfine Coupling. Simulation of the HYSCORE spectra assumes perfect C3h symmetry and one paramagnetic species and thus requires summing the hyperfine matrix contributions from the five carbon atoms from a Cp ring. The hyperfine matrix Ai for each Cp carbon atom i was modeled by point-dipolar contributions (Tijdd) due to spin density on Yb and the four adjacent carbons of the Cp ring, an isotropic contribution aiiso due to s-orbital spin density, and an axial contribution Tip due to π-orbital spin density.

A itotal ¼

5

∑ Tijdd þ aiiso I þ Tip j¼1

where bold symbols refer to matrices, italics to scalars, and I is the identity matrix. Each of the five point-dipolar matrices Tijdd are given (in Hz) by ij

Tdd ¼ Fj

T μ0 ge βe gn βn 3nij nij  1 rij3 2h

where nij and rij are respectively the unit column vector and the distance joining carbon i to the jth point of spin density at Yb or one of the four adjacent Cp ring carbons. Fj is the total spin density associated with nucleus j. Contributions to the hyperfine coupling from spin density on adjacent Cp rings are negligibly small due to the 1/r3 dependence and their small spin density. Due to the large spinorbit contribution from the Yb ion, this point-dipolar contribution was corrected with Tso dd = Tdd + ΔgTdd/ge, where Δg = g  geI. The spinorbit contribution for the spin density on the Cp ring is negligibly small. The spin density in the ligand-centered a0 MO is not uniformly distributed (Figure 10b), being concentrated on C2, C3, and C5. Values obtained from a DFT calculation are listed in section S9 of the Supporting Information. While we have little confidence in their absolute magnitude, the distribution of spin density within each ring should be reliable, because it is primarily determined by the aromatic π-interactions. Thus, the DFT Cp carbon s- and πorbital spin densities were used to provide a fixed set of scaling factors for the isotropic (nis) and anisotropic (nip) hyperfine coupling contributions for the carbon atoms, which are given by aiiso = nisAs and Tip = nipAip. The coefficients nis and nip are shown in Table S9 of the Supporting Information. The matrices Tip have principal values nip(Ap, Ap, 2Ap), with the unique axes pointing perpendicular to the C0 CiC00 planes. As and Ap are the only adjustable parameters and are used to match the simulated to the experimental HYSCORE spectra. The experimental s- and π-orbital spin densities for each carbon atom are calculated from the ratio of the experimental to the DFT hyperfine coupling multiplied by the corresponding DFT spin density, e.g. Fis,exp = (aiiso,exp/aiiso,DFT)Fis,DFT (cf. section S9 of the Supporting Information). The simulated spectra in Figure 3b use As = 0.4 ( 0.1 MHz and Ap = 1.6 ( 0.1 MHz. Table 6 lists the spin densities derived from these values.

Much weaker ridges in the spectra, corresponding to larger values of As and Ap, can be identified at lower contour levels (Figure S9.3 of the Supporting Information). We assign these to the minority species identified in the CW EPR spectrum (section 3.2.1); they have larger 13C spin densities, so we speculate that they may contain one or more η3-Cp rings. A ring orientation, in which an interacting allylic frontier MO has a nodal plane parallel to the nominal C3 axis, would reduce the steric strain around the Yb atom, while concentrating an enhanced spin density on only two carbon atoms, as opposed to the more general distribution of the η5 ring shown in Table 6. The DFT calculation attributes a negative spin density of 2.3% to the five ring hydrogen atoms, and when scaled by the factor required to reproduce the experimental densities at carbon, this becomes 0.11% per ring. The net spin density associated with all three Cp ligands is then 12.6 ( 0.9%. 4.6. Visible/Near-IR Spectrum. 4.6.1. Introduction. In 1967 Pappalardo and Jørgensen made a detailed spectroscopic study of thin sublimed films of YbCp3 at nominal temperatures of 4.2, 77, and 298 K.8 Noting a correlation between the energies of analogous broad transitions in four LnCp3 compounds and their reduction potentials, they assigned the absorption near 16,000 cm1 in YbCp3 to a ligand-to-metal charge transfer excitation. In TmCp3, for example, a similar transition is found at 23,800 cm1, consistent with a reduction potential for Tm3+ that is more negative by ∼1.2 eV. Although little further interpretation was possible in 1967, this data has proved to be invaluable. Compared to conventional “ionic” Yb3+ compounds, the spectrum of YbCp3 in the ff region is highly anomalous, spanning an energy range that is an order of magnitude larger and between 20 to 50 times more intense. (Figure 4 and Table 3). We show in section 3.4 that almost all of the intensity between 9,000 and 20,000 cm1 is vibronically “borrowed” from higher energy electric-dipole allowed transitions. The latter must be Cp f Yb charge-transfer excitations, because one-center excitations on Yb3+ (4f f 5d, at ∼5.5 eV)56 and Cp (π f π*, at ∼5.6 eV)57 are too high in energy to be significant. In C3h the a0 f a0 pure electronic transitions responsible for the visible absorption near 16,000 cm1 are electric-dipole forbidden, whereas a00 f a0 and e0 f a0 charge transfer transitions, polarized respectively parallel and perpendicular to the principal axis, are allowed. We tentatively assign two broad and much more intense bands, with maxima at 23,530 and 26,060 cm1, to these allowed transitions (section 3.4.7). They are respectively 1.0 and 1.30 eV higher in energy than the a0 f a0 transitions—intervals that agree well with the 1.3 eV difference between the ionization energy of the a0 ligand MO and those attributed to the e0 , e00 , and a00 MOs (cf. Figure 10a). 4.6.2. Selection Rules and Intensities. In the C3h* double group the electric-dipole operator transforms as Γ4(z) and z Γ2,Γ3(x ( iy), and the allowed transitions are Γ7,Γ8 f Γ9,Γ10 x ( iy

and Γ7,Γ8 f Γ9,Γ10 + Γ11,Γ12. In the model used to create Figure 11, the four excited states in the visible have the following symmetries: IV, Γ7,Γ8(forbidden); V, Γ11,Γ12 (allowed ^); VI, Γ9,Γ10 (allowed ^ and ); VII, Γ7,Γ8 (forbidden). Their intensities support this assignment. The total vibronic intensity on origins IV and VII is ∼30100 times more intense than that in the origin bands (Table 3), while origins V and VI have intensities that are similar in magnitude to that of the integrated vibronic structure built on them. Origin V at 11,385 cm1 in the solution spectrum is particularly strong and sharp (cf. Figure 6). )

Table 6. Experimental Spin Densities (%) on Ring Carbon Atoms

20655

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society Confirmation comes from the spectrum of the pyrrolidine adduct. The removal of the horizontal plane in the flattened tetrahedral geometry of this compound introduces a 10-fold increase in the intensity of the nominally forbidden transition to IV (section 3.4.5 and Table 4) and a 50% increase in the total band area based on the forbidden transition to origin VII, most of which is concentrated near the origin (section 3.4.6 and Table 4). There is however little intensity change in the region of origins V and VI, indicating that these pure electronic transitions are weakly electric-dipole allowed in the C3h group. In YbCp3 their intensity is unchanged in solution, while that in origin IV, attributable to the small out-of-plane displacement of Yb in the solid state, vanishes (Figure 5b) due to the presence of the C3h plane. ^

The intensity of the Γ7,Γ8 f Γ11,Γ12 (V) transition is of particular interest. Although it is magnetic-dipole forbidden, the origin is strong (fosc ∼ 2  105) compared to a typical 2F7/2 f 2 F5/2 transition (fosc ∼ 1  106) (Figure 4 and Tables 3 and 4), so its intensity must be attributed to CI. In the model of section 4.3.4, any CI is confined to the Γ7,Γ8 states (with orbital symmetry a0 ), while the basis of the Γ11,Γ12 states is pure 4f. It follows that the source of the intensity is an a0 f e0 orbital excitation (in the hole representation), with orbital transition moments of the form Æa0 |e(x ( iy)|4f(1æ. The dipole length in a two-center transition can be approximated by the distance of the (small) region of overlap density from the carbon nucleus, multiplied by the overlap integral SML. From the Æræ and Æ1/ræ for the Yb 4f5/2 and C (2p) orbitals,58 we estimate that this region is centered ∼165 pm from the carbon nucleus. However, the ligand contribution to the ground state is such that the c1 = 0.35, so that the effective dipole length is ∼58 pm. The experimental oscillator strength gives a dipole length of 0.75 pm, from which we estimate SML = 0.013—a value that matches those in section 4.1.3. This small overlap integral might appear incompatible with the relatively intense e0 f a0 transition in the 23,000 to 25,000 cm1 region, in which the a0 hole, predominantly centered on the Yb atom, is filled from the e0 ligand MO. However, the latter orbital, while primarily composed of the same Cp(π) orbitals as the a0 ligand-centered MO, also contains substantial 5d(2 character, so that the transition moments contain a one-center term of the form Æ5d(2|e(x ( iy)|4f(3æ that is intrinsically large. A closely related dipole length for the f14(1S0) f f13d(1P1) transition of Yb2+ ion was computed by Piper et al.,56 using the wave functions of Herman and Skillman59 and found to be 133 pm. The experimental intensity, section 3.4.7, gives a very approximate value for the dipole length in these transitions of ∼20 pm. Our DFT analysis (section S5 of the Supporting Information) gives the e0 ligand MO 5.8% 5d character, which, on the basis of Piper’s calculation, would predict a dipole length of 32 pm, and is in fair agreement with the experimental value. There is thus a consistency between the assignment and characteristics of excited states VIII or IX and the theoretical description of the composition of the Cp-π bonding MOs. On the basis of the theoretical orbital energies (section S5 of the Supporting Information), there should be an a00 f a0 charge transfer transition in the same spectral region as e0 f a0 . However, SML is too small for the two-center contributions to the transition moment to account for the experimental intensity, and we therefore look for one-center contributions. The only significant Yb contribution to the a00 orbital is 6pz (4.3%), but the a0 hole has no Yb contributions other than 4f(3 (and to a

ARTICLE

much lesser extent 4f(2), so no intensity can attributed to this source. Consideration of the nodal structure of the ligandcentered parts of the a00 and a0 orbitals similarly indicates that they also cannot contribute to the intensity. The presence of a second intense transition at an energy of ∼3.0 V cannot therefore be readily assigned to this excitation. An alternative is suggested by the presence of a vacant orbital, with the composition of a 6s/ 5d0 hybrid, 2.5 eV above the 4f shell, that is 1.2 eV more stable than any other virtual orbital. The conclusion to be drawn from the above discussion is that among the intense transitions near 3 eV only the e0 f a0 charge transfer excitation, polarized perpendicular to the C3 axis, is a plausible source of vibronic intensity in the near-infrared transitions. 4.6.3. Energies. The most striking feature of bands IV, V, and VI, nominally associated with the three components of the 2F5/2 state of the Yb3+ ion, is that they span 1997 cm1. A survey of 24 different Yb3+ compounds (section S7 of the Supporting Information) shows that, on average, these components span 597 ( 185 cm1 , with maximum and minimum values of 268 and 962 cm1. In the YbCp3 3 pyrrolidine adduct, the span is also anomalously large at 1080 cm1, but roughly half that in pure YbCp3. These observations can be understood by reference to Figure 11. When δCI = 0, the axial crystal field splits the 2 F5/2 state by 580 cm1, but when δCI = 3900 cm1, the splitting between origins IV, V, VI, and VII increases sharply so that their energies approximately match those that are observed (green arrows in Figure 11). The shift in the energy of these transitions between the solution and solid phase implies that δCI is significantly larger in solution (Figure 5). For example, the separation between origins IV and V, VI increases by ∼80 cm1. An increase in δCI in the solution phase is consistent with the observations in section 4.3.4 concerning the solution phase g-values and the solid state magnetic moment. As a result of the CI and the mixture of configurations in the ground state, some excitations, namely those to states V and VI, result in a net Yb f Cp charge-transfer, while those to states IV and VII involve a much larger transfer of charge in the opposite sense, i.e. Cp f Yb. Donation from a pyrrolidine ligand raises the energy of the 4f shell relative to Cp(π), increasing the energy of the “chargetransfer” band near 16,000 cm1 by ∼800 cm1. It also reduces the positive charge on the Yb atom, decreasing the 5d covalency and increasing the YbC bond length (experimentally by ∼5 pm; cf. section 3.3.1). Both factors reduce δCI, resulting in a decrease (Figure 11) in the splitting between bands IV, V, and VI, as observed. These results can be compared with those from Amberger’s group on tris(bis(trimethylsilyl)amido)ytterbium(III), or Yb(NSi2)3.60 These ligands are also strong donors, although the SiNSi planes are canted relative to the C3 axis, to give an effective site symmetry of D3 in solution, and the ligand field interaction differs from the YbCp3 case due to the presence of nitrogen orbitals having π-symmetry with respect to the YbN bond. In the solid state there is disorder within the nominal D3 site such that the Yb atoms lie above or below the plane of the amido groups by 52 pm to give a flattened trigonal pyramid local geometry, that is treated as having C3v symmetry.60,61 In this case, the charge-transfer transition is found at 27,250 cm1 (3.38 eV). The linearly polarized single crystal spectrum at 100 K shows three origins in the 2F5/2 region, all of which are electric-dipole 20656

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

Figure 13. Selected internal modes of the Cp ion.

allowed and which together span 1380 cm1. Their polarizations confirm that the central band at 11,176 cm1 has Γ5,6 symmetry in C3v*, which shares a basis with the C3h* Γ11,12 representation of origin V in YbCp3. Within our model the increased charge-transfer energy in Yb(NSi2)3 would be expected to reduce substantially the contributions from the charge-transfer configuration, one feature of which is a reduced energy span of the states with 2F5/2 parentage. Amberger et al.60 employ an empirical Hamiltonian limited to the Ψ0 4f13 basis, so their parameters should not be compared to those obtained here. 4.6.4. Vibronic Structure. The prominent modes in the vibronic spectrum either are internal to the Cp ligand with hν > 600 cm1 or are associated with the YbCp3 skeleton with hν < 300 cm1. Three normal modes of the Cp ligand at 1356, 1065, and 846 cm1 are readily assigned to ν10, ν11, and ν12 (in the notation of Bencze37) and transform as E20 in D5h. The first of these, nominally a CC stretching mode, is shown in Figure 13, in which σh represents the horizontal plane in YbCp3 and σv0 (not an operation of C3h) represents a nodal plane in the molecular a0 orbital (cf. Figure 10). The ν11 and ν12 modes are nominally CCH and CCC in plane bends, but all three normal modes are mixtures of these primitive displacements. In the charge-transfer configuration there is a distributed hole in the a0 MO. This orbital is bonding between C2C3, and to a lesser extent between C5C1 and C5C4, and also antibonding between C1C2 and C3C4. Optical excitations that modify the extent of the CI should therefore excite ν10, and to a lesser extent ν11 and ν12. Comparable linear vibronic coupling occurs in the photoelectron spectrum of the Cp anion,62 although this is complicated by a linear JahnTeller interaction along the same e20 normal coordinate, due to the electronic degeneracy in the ground state of the Cp radical. This degeneracy is absent in the charge transfer configuration of YbCp3 due to the large energy difference between the a0 and a00 Cp(π) MOs, but the linear electronphonon coupling should be comparable. Lineberger et al. analyzed the photoelectron spectrum of Cp• and report theoretical coupling constants of 0.21, 0.11, and 0.06 eV for ν10, ν11, and ν12, respectively,62 which is consistent with the trend in the intensities of the sidebands in Figure 7c. The

ARTICLE

coupling constants for the CH stretching modes are insignificant, and that for the a10 symmetrical ring breathing mode is also small.62 Consistent with these results, the latter modes do not appear in the optical spectrum of YbCp3. The 619 cm1 ν14 out-of-plane bending mode, also a source of significant vibronic intensity, transforms as E200 in D5h, and one of its components (Figure 13) folds the ring about the σv0 electronic nodal plane. Charge transfer from the Cp(π) MO selectively removes electronic charge density from C2, C3, and C5, while leaving that on C1 and C4 almost unchanged. Given that a Yb atom with a net charge of ∼2+ is located over the center of the ring, the ν14 folding displacement is the expected response to the nonuniformity in the ring carbon charges arising from the charge transfer. In YbCp3 each of the four vibronically active local ring modes combine to give collective molecular modes of a0 and e0 symmetry. Only the latter provide a source of electric dipole intensity, which must be drawn from an allowed a0 f e0 electronic transition. The obvious source is one of the intense electric-dipole allowed transitions to states VIII or IX (section 3.4.7), which lie ∼1.01.3 eV to higher energy and which is readily mixed with the a0 states by the electronphonon interaction, provided that an appropriate choice of phase is taken for the three constituent local ring displacements. The important point here is that the electronphonon interaction responsible for this intensity is associated exclusively with the ligands orbitals, so that its presence is a direct indication of the extent of CI in the initial and final states of this group of “forbidden” electronic transitions. Origins V and VI only exhibit weak vibronic structure attributable to internal ligand modes, while other origins, notably IV, show equally strong structure attributable to low frequency modes involving ring displacements relative to the Yb atom. The electric-dipole transition moment μ for a vibronic transition from the ground state |gæ to an electronic excited state |iæ in which one quantum of a vibration with frequency ν is also excited is given by Ægjμjiæ ¼

  1 p ÆgjqjkæÆkjμjiæ δE 2ν gk k   1 p ÆgjμjkæÆkjqjiæ  k δEik 2ν





ð6Þ

where q is the electronphonon coupling operator, |kæ is an excited state that is electric dipole allowed from either |gæ or |iæ, and δEgk and δEik are the relevant energy differences. We now make two assumptions. (a) There is only one significant state |kæ, namely that responsible for the intense absorption with energy δEgk ∼ 3 eV. For states for which i = IVVII, δEik is the range 1.751.00 eV. In section 4.6.2 we attributed the intensity near 3.0 eV to one-center transition moments of the form Æ5d(2|e(x ( iy)|4f(3æ, but analogous contributions of the type Æ5d(2|e(x ( iy)|4f(1æ will ensure that the transition moments Æk|μ|iæ are also significant. (b) States V and VI do not belong to the Γ7,Γ8 irreps whose states exhibit characteristics of CI, and they can therefore be attributed pure f13 character. The electronphonon coupling of a typical lanthanide state is expected to be too small to account for the substantial vibronic oscillator 20657

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society associated with these excited states, and we therefore write Æk|q|iæ ≈ 0 when i = V or VI. It follows that the vibronic intensity associated with origins V or VI originates in the first term in eq 6, is proportional to the square of the CI coefficient c1 in the ground state, and is inversely proportional to δEgk2. By contrast, in the case of origin IV, both terms in eq 6 are significant. The first term is similar to that just discussed, but the second term now contains the electron phonon coupling matrix element ÆVIII|q|IVæ. If the coupling is effective only on the ligand centered component of IV, represented by the CI coefficient c1(IV), this coupling is expected, using the eigenvectors from the calculation in section 4.3.4, to be larger than that in the ground state by the factor (c1(IV)/c1)(δEgk/δEik) = 2.52. The intensity derived in this way should therefore be six times more intense than that on origin V. In Table 4 the experimental ratio is 5.5 in the solid state and 8.0 in solution. This analysis can be extended to origin VII, whose vibronic transition moment relative to that on origin IV is given by (c1(VII)/c1(IV))(δE(VIIIIV)/δE(VIIIVII)) = 2.62, so that the predicted ratio of the oscillator strengths in these two transitions is 6.9. The experimental ratio (Table 4) in solution is 6.2 (the value of 4.4 in the solid is less reliable due to the large baseline corrections that have been applied in band VII). Three YbCp3 skeletal modes, at 138, 211, and 267 cm1, are prominent on origin IV. We associated these with IR and Raman active modes at 112, 196, and 255 cm1 (Table 2 and Table S2 of the Supporting Information), all of which have E0 symmetry, and are respectively a CpYbCp in plane bend, an asymmetric combination of YbCp tilting motions, and the asymmetric YbCp stretch. It is easy to understand how the displacement relative to the Yb3+ ion in these modes might mix the electricdipole forbidden a0 excited state, in which the ligand hole is uniformly distributed over all three rings, with the higher energy dipole-allowed e0 states, in which the hole amplitude is concentrated on either the x or y axes, in a configuration that would be modulated by the motion of the Yb3+ ion along either of these axes. The 138 cm1 CpYbCp bending mode appears to be the most strongly coupled of the three. To summarize, the vibronic structure in the electronic spectrum is consistent with, and provides additional evidence for, the presence of the configuration interaction, as discussed above. 4.7. Configuration Interaction and Covalency. The energies, intensities, and vibronic structure in the electronic transitions all support the predictions of the CI model. In particular, the transition energies serve to constrain the values of the CI matrix element, δCI = 3900 ( 400 cm1 and the energy difference between the configurations as ΔEML = 6100 ( 600 cm1. The wide error limits in these quantities arise from a significant correlation between them. The magnitude of δCI is conveniently discussed using the Superposition Model that has been used for the ab initio calculation of lanthanide CF parameters.44,46 This is closely related to the Angular Overlap Model,6365 in that the influence of each ligand is treated as additive. The CF parameters are resolved into radial and angular contributions—a separation that allows the overall ligand field to be expressed in terms of the properties of one or more cylindrically symmetrical metal ligand bonds. When applied to fluorides and oxides, the interactions include the ligand 2s, 2pσ, and 2pπ valence orbitals. Here, as a result of the bonding within the Cp ligand, we assume that the YbC overlap is confined to those C2p atomic orbitals that

ARTICLE

Table 7. Charge Transfer Energies and CI Coefficients host

E(CT)/eV

c21

CaF2

8.61

0.002

YbF63

7.35

0.004

LiScO2

6.02

0.005

Y3Al5O12

5.79

0.005

Y2O3

5.46

0.006

[LnCl6]3

4.55

0.008

[LnBr6]3

3.63

0.012

YbCp3

1.24

0.11

constitute the Cp(2pπ) MOs and is therefore comparable to the Yb2pσ interaction in these much-studied “ionic” examples. An ab initio calculation of the CF parameters arising from PrF and PrCl bonds has been described by Newman.46 The contributions to the CF interaction fall into two groups: (a) those associated with a pure 4f electron Slater-determinant, such as Ψ0 in eq 1, and (b) those associated with covalency represented, as in the current case, by the contribution of the chargetransfer configuration Ψ1. The former group includes the classical electrostatic 4f electron ligand point-charge interaction while also allowing for the penetration of 4f charge into the distributed electronic charge on the ligand, as well as overlap and exchange contributions, associated with the orthogonalization of the 4f and ligand valence shells that arise from the Pauli exclusion between them. The covalent component is determined by Ψ1, weighted by the coefficient c1 = Nfσ/Dfσ. When c1 is small, the denominator Dfσ can be identified with the energy of the charge-transfer transition between the two configurations. The various contributions to Nfσ are linearly dependent on the overlap integral Æ4f|σæ. In PrF3, by far the largest of these is the interaction of the exchange-charge (i.e., the 4f2p(σ) overlap density) with the ligand nuclear and electronic charges.46,47 Given that the overlap integral is small, it is perhaps surprising that in PrF3, at a distance of 4.99 Bohr, Nfσ is calculated to be 3273 cm1 or 0.41 eV. However, integrals of similar magnitude have been calculated for the Sm2+F interaction in samarium doped BaFCl.44 A recent and more sophisticated calculation for Yb3+ doped into Cs2NaYF6, which yields excellent agreement with the experimental 19F hyperfine coupling constants, gives Nfσ = 0.458 eV (3700 cm1).66 In the absence of appropriate computational tools, we now assume that Nfσ is of comparable magnitude in YbCp3, as is indeed suggested by the configuration interaction matrix element of 3900 cm1 (0.48 eV) derived empirically in this work. Values of Dfσ can be deduced from the optical charge transfer energies for Yb3+ in various hosts and complexes67,68 and are listed in Table 7, which also lists c21 based on the obviously crude assumption that Nfσ remains constant at 0.41 eV. In YbCp3 the equivalent value of Dfσ, after taking into account the spinorbit interaction in the 4f13 state, is 1.24 eV. The conclusion to be drawn from Table 7 is that covalency, as measured by c21, scales as the square of 1/E(CT) and so plays only a minor part in determining the CF interactions in “ionic” compounds, but it rapidly becomes important when the orbital energies on the ligand approach those of the 4f shell in Yb3+. There appears therefore to be nothing anomalous about the YbC overlap or the CI matrix element that depends on it. The significant covalent contribution in YbCp3 is mainly determined 20658

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society

ARTICLE

Table 8. Square of the CI Coefficient by Various Methods c21

observable PES intensities

0.120

g-tensor principal values

0.124

171

0.17 ( 0.03

13

0.126 ( 0.009

Yb hyperfine coupling

C hyperfine coupling

by the very low energy of the ligand to metal charge-transfer transition, which is a consequence of the strong donor properties of the Cp ligand and the low reduction potential of Yb3+.

5. CONCLUSION This work seeks to quantify the extent of covalency in the 4f shell of Yb3+, in terms of the contribution of the Cp f Yb3+ charge-transfer configuration in the ground state wave function. Table 8 summarizes values for the square of the configuration interaction (CI) coefficient c1, obtained by four independent methods. The two configurational components of the ground state give rise to distinct photoionization spectra at high photon energy, whose relative intensities allow a direct determination of c1. The corollary of Cp f Yb charge-transfer is Yb f Cp spin transfer. The reduced spin-density on the ytterbium atom is revealed in the EPR experiment by a reduction in the 171Yb hyperfine coupling, while the 13C hyperfine interactions, as determined in the HYSCORE experiment, quantify the increase in spin density on the cyclopentadienyl carbon atoms. Each experiment provides a separate measure of c1. Finally, the axial anisotropy in the g-tensor can also be used to determine this coefficient, because the axial component of the total angular momentum, nominally maintained by a very large spinorbit interaction, can be effectively partially quenched through the participation of the charge-transfer configuration. The highly anomalous energies, intensities, and vibronic structure in the electronic absorption spectrum also provide a strong indication of the presence of the charge-transfer configuration. Taken with the g-value anisotropy, the ff excitation energies also constrain the magnitude of c1. From an analysis of the factors that control the CI coefficient, it appears that the absence of significant covalency in the 4f shell of conventional compounds should not be attributed to small overlap integrals, but rather to the large values of ECT (Table 8) that are an inevitable consequence of the electropositivity of the lanthanides and the electronegativity of their common ligands. For strongly donating electron-rich ligands, a substantial degree of 4f shell covalency is to be expected when the lanthanide(III) ions are readily reducible. For example, relative to Yb(III), the change in ECT, averaged over a large number of III compounds, is 0.44, +0.72, and +1.23 eV for Eu, Sm, and Tm, respectively, but ∼+2 eV for Pr, Nd, Dy, Ho, and Er.67 In this latter group, any 4f covalency in LnCp3 compounds should therefore be small (c21 e 1.6%). It will, however, be of major significance in EuCp3 and, though smaller than in YbCp3, should also be easily identifiable in SmCp3 and TmCp3 (c21 ≈ 34%). Clearly, a small increase in ligand basicity should greatly enhance this element of the covalency in ytterbium organometallics. Notice, however, that the proximity on the energy scale of a ligand frontier orbital that interacts strongly with the 4f shell, in our example a0 (f3c) (Figure 10a), is a consequence of the absence of its interaction with the 5d shell and arises from the C3h

symmetry. The generality of this type of covalency will therefore be limited, by the presence of 5d ligand orbital interactions, to cases where local symmetry constraints allow the ligand frontier orbital energies to approach those of the 4f shell. Finally, we emphasize that our results only shed light on the characteristics of the 4f shell. They do not directly quantify bonding interactions in the 5d shell, and they should not be taken to imply that the bonding in lanthanide cyclopentadienyls is primarily “ionic”.

’ ASSOCIATED CONTENT

bS

Supporting Information. Crystallographic data (CIF) file, selected bond metric data, vibrational frequencies, EPR and HYSCORE data, a list of published Yb3+ EPR parameters, computational geometry optimizations, orbital energies and spin densities, and a list of published Yb3+ optical excitation energies, as well as a full list of authors for ref 28. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

[email protected]

’ ACKNOWLEDGMENT We thank Dr. Andrea Sella and Charlene Hunston of University College, London for providing samples of YbCp3. The University of Oxford provided crystallographic services and a DTA studentship for M.I. The Engineering and Physical Sciences Research Council supported the Oxford Center for Advanced Electron Spin Resonance (Grant EP/D044855D/1) and also provided financial support for M.I. ’ REFERENCES (1) Coreno, M.; de Simone, M.; Green, J. C.; Kaltsoyannis, N.; Narband, N.; Sella, A. Chem. Phys. Lett. 2006, 432, 17. (2) Coreno, M.; de Simone, M.; Coates, R.; Denning, M. S.; Denning, R. G.; Green, J. C.; Hunston, C.; Kaltsoyannis, N.; Sella, A. Organometallics 2010, 29, 4752. (3) Coates, R.; Coreno, M.; DeSimone, M.; Green, J. C.; Kaltsoyannis, N.; Kerridge, A.; Narband, N.; Sella, A. Dalton Trans. 2009, 5943. (4) Bond, A. M.; Deacon, G. B.; Newnham, R. H. Organometallics 1986, 5, 2312. (5) Booth, C. H.; Walter, M. D.; Daniel, M.; Lukens, W. W.; Andersen, R. A. Phys. Rev. Lett. 2005, 95. (6) Booth, C. H.; Kazhdan, D.; Werkema, E. L.; Walter, M. D.; Lukens, W. W.; Bauer, E. D.; Hu, Y. J.; Maron, L.; Eisenstein, O.; HeadGordon, M.; Andersen, R. A. J. Am. Chem. Soc. 2010, 132, 17537. (7) Booth, C. H.; Walter, M. D.; Kazhdan, D.; Hu, Y. J.; Lukens, W. W.; Bauer, E. D.; Maron, L.; Eisenstein, O.; Andersen, R. A. J. Am. Chem. Soc. 2009, 131, 6480. (8) Pappalardo, R.; Jørgensen, C. K. J. Chem. Phys. 1967, 46, 632. (9) Schlesener, C. J.; Ellis, A. B. Organometallics 1983, 2, 529. (10) Cosier, J.; Glazer, A. M. J. Appl. Crystallogr. 1986, 19, 105. (11) Otwinowski, Z.; Minor, W. Methods Enzymology; Academic Press: New York, 1997. (12) Sheldrick, G. M. Acta Crystallogr., Sect. A 2008, 64, 112. (13) Sheldrick, G. M. Institut fur Anorganische Chemie der Universitat: Gottingen, Germany, 1998. (14) Calle, C.; Sreekanth, A.; Fedin, M. V.; Forrer, J.; Garcia-Rubio, I.; Gromov, I. A.; Hinderberger, D.; Kasumaj, B.; Leger, P.; Mancosu, B.; 20659

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660

Journal of the American Chemical Society Mitrikas, G.; Santangelo, M. G.; Stoll, S.; Schweiger, A.; Tschaggelar, R.; Harmer, J. Helv. Chim. Acta 2006, 89, 2495. (15) Schweiger, A.; Jeschke, G. Principles of pulse electron paramagnetic resonance; Oxford University Press: Oxford, U.K.; New York, 2001. (16) Stoll, S.; Schweiger, A. J. Magn. Reson. 2006, 178, 42. (17) ADF2004.01 ADF2010.02, S., Theoretical Chemistry; Vrije Universiteit: Amsterdam, The Netherlands, 2009; http://www.scm. com. (18) te Velde, G.; Bickelhaupt, F. M.; Baerends, E. J.; Fonseca Guerra, C.; Van Gisbergen, S. J. A.; Snijders, J. G.; Ziegler, T. J. Comput. Chem. 2001, 22, 931. (19) Fonseca Guerra, C.; Snijders, J. G.; te Velde, G.; Baerends, E. J. Theor. Chem. Acc. 1998, 99, 391. (20) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1993, 99, 4597. (21) van Lenthe, E.; Baerends, E. J.; Snijders, J. G. J. Chem. Phys. 1994, 101, 9783. (22) van Lenthe, E.; Snijders, J. G.; Baerends, E. J. J. Chem. Phys. 1996, 105, 6505. (23) van Lenthe, E.; van Leeuwen, R.; Baerends, E. J.; Snijders, J. G. Int. J. Quantum Chem. 1996, 57, 281. (24) Vosko, S. H.; Wilk, L.; Nusair, M. Can. J. Phys. 1980, 58, 1200. (25) Becke, A. D. Phys. Rev. A 1988, 38, 3098. (26) Becke, A. D. J. Chem. Phys. 1988, 88, 1053. (27) Perdew, J. P. Phys. Rev. B 1986, 33, 8800. (28) Frisch, M. J.; et al. Gaussian 03, rev. D.02; Wallingford, CT, 2004. (29) Stephens, P. J.; Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J. J. Phys. Chem. 1994, 98, 11623. (30) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785. (31) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (32) Eggers, S. H.; Kopf, J.; Fischer, E. O. Acta Crystallogr., Sect. C: Cryst. Struct. Commun. 1987, 43, 2288. (33) Baker, E. C.; Raymond, K. N. Inorg. Chem. 1977, 16, 2710. (34) Deacon, G. B.; Fallon, G. D.; Forsyth, C. M.; Gatehouse, B. M.; Junk, P. C.; Philosof, A.; White, P. A. J. Organomet. Chem. 1998, 565, 201. (35) Baisch, U.; Pagano, G.; Zeuner, M.; Barros, N.; Maron, L.; Schnick, W. Chem.—Eur. J. 2006, 12, 4785. (36) Ni, C. H.; Deng, D. O.; Qian, C. T. Inorg. Chim. Acta—F-Block Elem. Art. Lett. 1985, 110, L7. (37) Bencze, E.; Lokshin, B. V.; Mink, J.; Herrmann, W. A.; Kuhn, F. E. J. Organomet. Chem. 2001, 627, 55. (38) Aleksanyan, V. T.; Borisov, G. K.; Devyatych, G. G.; Gachter, B. F.; Koningstein, J. A.; Schneider, B. E. J. Raman Spectrosc. 1974, 2, 345. (39) Reinhard, C.; Gudel, H. U. Inorg. Chem. 2002, 41, 1048. (40) Deloach, L. D.; Payne, S. A.; Chase, L. L.; Smith, L. K.; Kway, W. L.; Krupke, W. F. IEEE J. Quantum Electron. 1993, 29, 1179. (41) Carnall, W. T.; Fields, P. R.; Wybourne, B. G. J. Chem. Phys. 1965, 42, 3797. (42) Amberger, H. D.; Schultze, H. Spectrochim. Acta, Part A: Mol. Biomol. Spectrosc. 1987, 43, 1301. (43) Angelov, B. M. Chem. Phys. Lett. 1975, 31, 503. (44) Shen, Y. R.; Bray, K. L. Phys. Rev. B 1998, 58, 5305. (45) Pearson, J. J.; Herrmann, G. F. J. Appl. Phys. 1969, 40, 1142. (46) Newman, D. J. Adv. Phys. 1971, 20, 197. (47) Newman, D. J.; Curtis, M. M. J. Phys. Chem. Solids 1969, 30, 2731. (48) Hubbard, J.; Rimmer, D. E.; Hopgood, F. R. A. Proc. Phys. Soc. London 1966, 88, 13. (49) Koster, G. F.; Dimmock, J. O.; Wheeler, R. G.; Schatz, H. Properties of the Thirty-Two Point Groups; MIT Press: Cambridge, MA, 1963. (50) Wybourne, B. G. J. Chem. Phys. 1965, 43, 4506. (51) Morrison, C. A.; Leavitt, R. P. J. Chem. Phys. 1979, 71, 2366. (52) Zhou, X. J.; Reid, M. F.; Faucher, M. D.; Tanner, P. A. J. Phys. Chem. B 2006, 110, 14939. (53) Freeman, A. J.; Desclaux, J. P. J. Magn. Magn. Mater. 1979, 12, 11.

ARTICLE

(54) Electron Paramagnetic Resonance; Abragam, A., Bleaney, B., Eds.; Oxford: Clarendon Press: Oxford, 1970. (55) Dong, H. N.; Wu, S. Y.; Luo, X. B.; Tang, H. Spectrochim. Acta, Part A: Mol. Biomol. Spectrosc. 2005, 61, 637. (56) Piper, T. S.; Brown, J. P.; McClure, D. S. J. Chem. Phys. 1967, 46, 1353. (57) Wagner, B. O.; Ebel, H. F. Tetrahedron 1970, 26, 5155. (58) Desclaux, J. P. At. Nucl. Struct. Tables 1973, 12, 311. (59) Herman, F.; Skillman, S. Atomic Structure Calculations; Prentice-Hall, Inc.: Englewood Cliffs, NJ, 1963. (60) Amberger, H. D.; Jank, S.; Reddmann, H.; Apostolidis, C. Z. Anorg. Allg. Chem. 2007, 633, 398. (61) Niemeyer, M. Z. Anorg. Allg. Chem. 2002, 628, 647. (62) Lineberger, W. C.; Ichino, T.; Wren, S. W.; Vogelhuber, K. M.; Gianola, A. J.; Stanton, J. F. J. Chem. Phys. 2008, 129, 084310. (63) Gerloch, M.; Woolley, R. G. Prog. Inorg. Chem. 1983, 31, 371. (64) Schaffer, C. E.; Jørgensen, C. K. Mol. Phys. 1965, 9, 401. (65) Newman, D. J.; Ng, B. Rep. Prog. Phys. 1989, 52, 699. (66) Falin, M. L.; Anikeenok, O. A.; Latypov, V. A.; Khaidukov, N. M.; Callens, F.; Vrielinck, H.; Hoefstaetter, A. Phys. Rev. B 2009, 80. (67) Dorenbos, P. J. Phys.: Condens. Matter 2003, 15, 8417. (68) Anikeenok, O. A.; Eremin, M. V.; Falin, M. L.; Konkin, A. L.; Meiklyar, V. P. J. Phys. C: Solid State Phys. 1984, 17, 2813.

20660

dx.doi.org/10.1021/ja209311g |J. Am. Chem. Soc. 2011, 133, 20644–20660